Abstract
Depression has been associated with negatively biased cognitive styles, particularly negative self-attribution,
i.e. increased self-attribution of negative outcomes and/or decreased self-attribution of positive outcomes; and
negative self-evaluation, i.e. being underconfident. Since these (meta)cognitions concern one’s actions and
their outcomes, these may be relevant to understanding alterations in value-based learning. We investigated
this computationally, assessing how value-based learning may be influenced by different patterns in causal
attribution, or in confidence (for an irrelevant, perceptual task). I will discuss initial results from an online study
in the general population, combining adapted learning tasks with a transdiagnostic questionnaire battery,
centred on depressive and related mental health symptoms. Replicating previous findings, we found that selfattribution
patterns can shape learning biased. We also see some promise for this task in quantifying negative
self-attribution biases linked to depressive symptoms. In a second task, we see some evidence that value-based
learning can be disrupted by (even irrelevant) confidence, but neither this effect, nor average confidence, were
linked to depressive symptoms. Further ongoing model-based analyses will be discussed, seeking to
disentangle interactions between learning, self-attribution, and self-evaluation, and links to mental health.